-
-
Notifications
You must be signed in to change notification settings - Fork 80.7k
[Bug]: [Regression] 2026.5.27: Codex app-server turn-completion stall returns ("Codex stopped before confirming the turn was complete") — regression of #84076 (fixed by #85107) #88312
Copy link
Copy link
Closed
Labels
P1High-priority user-facing bug, regression, or broken workflow.High-priority user-facing bug, regression, or broken workflow.bugSomething isn't workingSomething isn't workingclawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.ClawSweeper found a clear likely implementation shape for this issue.clawsweeper:needs-live-reproClawSweeper needs live local, crabbox, or manual validation to confirm this issue.ClawSweeper needs live local, crabbox, or manual validation to confirm this issue.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:message-lossChannel message delivery can be lost, duplicated, or misrouted.Channel message delivery can be lost, duplicated, or misrouted.impact:session-stateSession, memory, transcript, context, or agent state can drift or corrupt.Session, memory, transcript, context, or agent state can drift or corrupt.issue-rating: 🐚 platinum hermitGood issue quality with a plausible reproduction path needing some confirmation.Good issue quality with a plausible reproduction path needing some confirmation.regressionBehavior that previously worked and now failsBehavior that previously worked and now fails
Description
Metadata
Metadata
Assignees
Labels
P1High-priority user-facing bug, regression, or broken workflow.High-priority user-facing bug, regression, or broken workflow.bugSomething isn't workingSomething isn't workingclawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.ClawSweeper found a clear likely implementation shape for this issue.clawsweeper:needs-live-reproClawSweeper needs live local, crabbox, or manual validation to confirm this issue.ClawSweeper needs live local, crabbox, or manual validation to confirm this issue.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:message-lossChannel message delivery can be lost, duplicated, or misrouted.Channel message delivery can be lost, duplicated, or misrouted.impact:session-stateSession, memory, transcript, context, or agent state can drift or corrupt.Session, memory, transcript, context, or agent state can drift or corrupt.issue-rating: 🐚 platinum hermitGood issue quality with a plausible reproduction path needing some confirmation.Good issue quality with a plausible reproduction path needing some confirmation.regressionBehavior that previously worked and now failsBehavior that previously worked and now fails
Type
Fields
Priority
None yet
Bug type
Regression (worked before, now fails)
Beta release blocker
No
Summary
On the Codex app-server (ChatGPT Plus sub), a multi-tool agent turn reliably fails with Codex stopped before confirming the turn was complete starting in 2026.5.27; 2026.5.26 completes the same turns. Identical symptom to #84076 (P1, closed by #85107) → that fix appears to have regressed in 2026.5.27, still broken in 2026.5.28-beta.*.
Bisect: 2026.5.26 ✅ (stopReason: stop, ~23 tool calls) → 2026.5.27 ❌ → 5.28-beta.1 ❌ → 5.28-beta.4 ❌. From production cron history: same job completed on 5.26, failed the first run after the 5.27 upgrade and every run since.
Ruled out: not model choice (5.4 & 5.5 both stall); not the ~05-26/27 OpenAI backend change (5.26 coped on 05-28 — tracks the OC version, not a date); idle-burn is a separate symptom (5.27 also regressed it; beta.1 fixed that but not turn-completion).
Steps to reproduce
openclaw agent --model openai/gpt-5.4 --message "<~20 sequential tool calls then summarize>" — completes on 5.26, stalls on 5.27+. (Self-contained repro script available on request.)
Expected behavior
stopReason: stop after a longish (23 tool calls) turn. Works on 5.26. Doesn't on 5.27 and 5.28?beta*
Actual behavior
Codex stops without saying the turn is complete or emitting an error.
OpenClaw version
2026.5.28-beta.4
Operating system
Debian Bookworm
Install method
npm global
Model
GPT-5.4
Provider / routing chain
openai/gpt-5.4 (via a Codex Pro sub)
Additional provider/model setup details
No response
Logs, screenshots, and evidence
Impact and severity
Affected: OC users, cron jobs, various handlers, etc.
Severity: Critical. OC unusable on Codex sub.
Frequency: Deterministic. All longish turns (multiple tool calls) reliably fail.
Only found workaround is pinning 5.26.
Additional information
Bisect: 2026.5.26 ✅ (stopReason: stop, ~23 tool calls) → 2026.5.27 ❌ → 5.28-beta.1 ❌ → 5.28-beta.4 ❌.
Only found workaround is pinning 5.26.